fix(agc): draw the complete rectangles of a rect-list draw with a partial last one

A rect-list draw whose count is not a multiple of 3 threw "incomplete rect-list
primitive". Rect lists are drawn as patch lists of 3 control points, and Vulkan
generates floor(vertexCount/3) patches, so the draw now records as is and the
trailing 1 or 2 vertices are left out. Palworld (PPSA23126) draws 4 vertices x 9
instances as a rect list when character creation opens (#3832).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

core/libs/prx/libSceAgcDriver/Graphics/src/Draw.cpp

@@ -1025,7 +1025,6 @@ DrawInputs prepareDrawInputs(const Context& context, const State& state, const P
     if (draw.indexed) GuestMemory::CheckRange(reinterpret_cast<const void*>(draw.indexAddress), static_cast<std::size_t>(indexBytes), draw.indexSize);
     APS5_LOG_CHARS_OUT_DEBUG("Index buffer range OK");
     Require(!draw.indexed || !state.hasColorTarget || draw.indexAddress + indexBytes <= state.color.address || state.color.address + state.color.bytes <= draw.indexAddress, "index buffer aliases the render target");
-    if (state.rectList) Require(draw.indexCount % 3 == 0, "incomplete rect-list primitive");
     APS5_LOG_CHARS_OUT_DEBUG("ValidateShaders");
     if (recipe != nullptr) {
         inputs.fragmentOutputs = recipe->fragmentOutputs;

core/libs/prx/libSceAgcDriver/tests/Graphics.cpp

@@ -3025,8 +3025,17 @@ void rectListTests() {
     queue.userConfig[0x242] = 17;
     const auto state = DecodeState(queue);
     const Context context{};
-    const AgcDriver::Pm4::DrawParameters draw{0, 4, 0, 1, 0, false};
-    expectFailure([&] { Draw(context, state, draw, {}); }, "incomplete rect-list");
+    const auto drawFailure = [&](std::uint32_t count) {
+        const AgcDriver::Pm4::DrawParameters draw{0, count, 0, 1, 0, false};
+        try {
+            Draw(context, state, draw, {});
+        } catch (const std::exception& error) {
+            return std::string(error.what());
+        }
+        return std::string{};
+    };
+    const auto complete = drawFailure(3);
+    for (const auto count : {4u, 5u}) Require(drawFailure(count) == complete, "a rect-list draw with a partial last rectangle was not drawn like its complete rectangles");
 }
 
 ShaderRecompiler::RecompileResult recompilePixel(std::initializer_list<std::uint32_t> controls, std::span<const std::uint32_t> code) {

docs/dev/TechnicalDebt.md

@@ -357,6 +357,7 @@ Throughout the project, every function at every stage either **does exactly what
 - [Windows file identity](../../core/libs/prx/libkernel/File/src/NativeStat.cpp) (libkernel) - `stat` and `fstat` report the volume serial number as `st_dev` and the 64-bit file index folded into the guest's 32-bit `st_ino` (never 0), so two different files can in principle share an `st_ino`; the fold is not taken from a reference. `st_nlink` is clamped to 65535, the width of the guest field. The call keeps the `_wstat64` values when `GetFileInformationByHandle` fails
 - On Linux, [direct memory](../../core/libs/prx/libkernel/DirectMemory/DirectMemory.cpp) lives in one sparse memfd at file offset = physical address, so the [GPU imports](../../core/libs/prx/libSceAgcDriver/Graphics/src/GuestBufferMemory.cpp) can share one udmabuf buffer object per 64 MiB of it. Released direct memory is not punched out of the file, because a live import of its chunk pins the pages and the GPU would keep reading the old ones: its pages stay committed (the process keeps the high-water mark of the direct memory it used) and are zeroed when the range is allocated again.
 - [Per-face depth bias](../../core/libs/prx/libSceAgcDriver/Graphics/src/Draw.cpp) (libSceAgcDriver): Vulkan has one depth bias per draw, so a draw that rasterizes both faces with different front and back bias (PA_SU_SC_MODE_CNTL, PA_SU_POLY_OFFSET_FRONT/BACK_*) is recorded twice, front faces with the back faces culled and then back faces with the front faces culled. The vertex stages run twice, so a vertex shader that writes memory writes twice, and all front faces land before all back faces instead of in primitive order, which differs only for blending or stencil updates that depend on the order.
+- [Partial rect-list primitives](../../core/libs/prx/libSceAgcDriver/Graphics/src/Draw.cpp) (libSceAgcDriver): a rect-list draw whose vertex or index count is not a multiple of 3 draws its complete rectangles and leaves out the 1 or 2 vertices after them, because rect lists are drawn as Vulkan patch lists of 3 control points, which generate [⌊vertexCount/3⌋ patches](https://registry.khronos.org/vulkan/specs/latest/html/vkspec.html#drawing-patch-lists). What the console draws for the incomplete rectangle is not measured; Palworld (PPSA23126) draws 4 vertices × 9 instances as a rect list when its character creation opens.
 - [Separate alpha blend into 8_8 SWAP_ALT targets](../../core/libs/prx/libSceAgcDriver/Graphics/src/Pipeline.cpp) (libSceAgcDriver): the second channel of a two-channel target with `COMP_SWAP` 1 holds the alpha export, but Vulkan blends it with the color equation of the R8G8 attachment, so a draw whose alpha equation differs from its color equation is recorded twice, the first channel with the color equation and the second with the alpha factors and operation moved into the color equation. The vertex stages run twice, so a vertex shader that writes memory writes twice, and occlusion counts include both passes. Draws that write depth or stencil, pixel shaders that write memory and factors of the destination alpha throw. That the hardware blends the second channel with the alpha equation is taken from Mesa's `si_translate_colorswap` and radeonsi's use of it for luminance-alpha formats, not measured on hardware.
 - [Compute LDS beyond the device's shared-memory limit](../../core/libs/prx/libSceAgcDriver/Execution/src/VulkanDevice.cpp) (libSceAgcDriver) lives in a coherent storage buffer with one slice of `LDS bytes / 4 + 1` dwords per workgroup (the last dword is the atomic lock), shared by all such dispatches and grown in 1 MiB steps up to `maxStorageBufferRange`; a dispatch whose slices do not fit throws. It is slower than shared memory and its cost was not measured. An indirect dispatch of such a program always reads its group counts on the CPU. Its results are pinned against the in-limit semantics in `agc_driver_lds_beyond_device_limit`, not measured on hardware.
 - [Graphics pipeline libraries](../../core/libs/prx/libSceAgcDriver/Graphics/src/PipelineLibrary.cpp) (libSceAgcDriver): a pipeline is fast-linked at the draw and relinked with `VK_PIPELINE_CREATE_LINK_TIME_OPTIMIZATION_BIT_EXT` on a background thread, and binds the optimized one from the next draw on. Draws recorded before that use the fast link, whose vertex stage keeps the outputs the fragment stage does not read: on RADV (Radeon 660M, Mesa 26.2.4) 346 of 483 Astro's Playroom vertex shaders paired with a fragment stage that reads none of them used more VGPRs (median 8 more, up to 2.5 times), and on an RTX 4090 13 of 15 PPSA36082 vertex shaders up to 68 instead of 38 (#2242). The optimized link compiles to the same code as a whole pipeline